View
11
Download
0
Category
Preview:
Citation preview
RESEARCH ARTICLE
How inter-state amity and animosity
complement migration networks to drive
refugee flows: A multi-layer network analysis,
1991–2016
Justin SchonID1*, Jeffrey C. Johnson2
1 University of Virginia, Charlottesville, Virginia, United States of America, 2 University of Florida,
Gainesville, Florida, United States of America
* jss5yf@virginia.edu
Abstract
What drives the formation and evolution of the global refugee flow network over time? Refu-
gee flows in particular are widely explained as the result of pursuits for physical security,
with recent research adding geopolitical considerations for why states accept refugees. We
refine these arguments and classify them into explanations of people following existing
migration networks and networks of inter-state amity and animosity. We also observe that
structural network interdependencies may bias models of migration flows generally and ref-
ugee flows specifically. To account for these dependencies, we use a dyadic hypothesis
testing method—Multiple Regression- Quadratic Assignment Procedure (MR-QAP). We
estimate MR-QAP models for each year during the 1991–2016 time period. K-means clus-
tering analysis with visualization supported by multi-dimensional scaling allows us to identify
categories of variables and years. We find support for the categorization of drivers of refu-
gee flows into migration networks and inter-state amity and animosity. This includes key
nuance that, while contiguity has maintained a positive influence on refugee flows, the mag-
nitude of that influence has declined over time. Strategic rivalry also has a positive influence
on refugee flows via dyad-level correlations and its effect on the structure of the global refu-
gee flow network. In addition, we find clear support for the global refugee flow network shift-
ing after the Arab Spring in 2011, and drivers of refugee flows shifting after 2012. Our
findings contribute to the study of refugee flows, international migration, alliance and rivalry
relationships, and the application of social network analysis to international relations.
Introduction
Refugee flows have become a source of global concern. Origin countries fear lost human capi-
tal, lost legitimacy for their governments, and the possibility that political opponents will estab-
lish bases outside their reach [1–3]. Destination countries fear the tensions that come with
shifts in their demographic balances and the diffusion of conflict and terrorism [4, 5]. Govern-
ments in many destination countries have therefore implemented measures to secure their
borders and prevent immigration [6, 7]. Yet, refugee flows may follow specific patterns due to
PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 1 / 13
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Schon J, Johnson JC (2021) How inter-
state amity and animosity complement migration
networks to drive refugee flows: A multi-layer
network analysis, 1991–2016. PLoS ONE 16(1):
e0245712. https://doi.org/10.1371/journal.
pone.0245712
Editor: Jordi Paniagua, Universitat de Valencia,
SPAIN
Received: July 17, 2020
Accepted: January 7, 2021
Published: January 27, 2021
Copyright: © 2021 Schon, Johnson. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
information files.
Funding: JJS and JJ acknowledge support from
the Army Research Office/Army Research
Laboratory. The research was supported in part by
the Army Research Office/Army Research
Laboratory under award no. W911NF1810267
(Multidisciplinary University Research Initiative).
The views and conclusions contained in this
manuscript are those of the author and should not
networks that exist within the international system [8, 9]. These networks may be too powerful
for individual countries to effectively block refugee flows. This leads to the question: What
drives the formation and evolution of the global refugee flow network over time?
To answer this question, we contend that it is valuable to examine the many overlapping
networks that form the international system [10–12]. Disparities in well-being are important,
but they do not tell the whole story [13]. Networks of international trade, international migra-
tion, alliances, rivalry, physical distance, and other factors combine to form the structure of
motivations and opportunities that drive global refugee flows [8, 14–19].
Our analysis focuses on the post-Cold War 1991–2016 time period. As migration research
such as Orchard [20] observes, the post-Cold War time period has experienced multiple phases
in refugee flow patterns. Large refugee flows during the early 1990s that resulted from episodes
such as the breakup of Yugoslavia, Siad Barre’s removal from power in Somalia, the Taliban’s
takeover of Afghanistan, and mass killing and genocide in Rwanda and Burundi were met with
the exhaustion of American and other Western governments from hosting refugees during the
Cold War [20]. Neighboring countries such as Kenya, Iran, Pakistan, Tanzania, and the Demo-
cratic Republic of the Congo welcomed refugees, but their hospitality progressively waned
by the end of the 1990s [21–23]. In the early 2000s, governments around the world rapidly
increased the amount of border walls and obstacles to refugee flows and international migration
[7, 24–26]. By the end of the 2000s, new refugee flows had fallen to relatively low levels. In 2011,
the Arab Spring sparked renewed surges of refugees. Governments fell and civil wars began
across Tunisia, Libya, Egypt, Syria, Yemen, and Iraq [27]. These developments destabilized
other countries as well. Libya’s collapse contributed to flows of weapons and fighters into Mali,
Niger, Chad, Sudan, and other countries in the African Sahel [28]. While these developments
primarily led to internal displacement in 2011, refugee flows increased substantially starting in
2012. The 2012–2016 time period marks the clearest coherent phase of refugee flows after the
Cold War, so we believe that this is a particularly valuable time period to analyze.
We hypothesize that inter-state geopolitical networks, specifically networks of amity and
networks of animosity, and international migration networks are the most important types of
factors for explaining refugee flows. While the growing field of refugee studies often attempts
to explain refugee flows, like international migration, through a focus on international migra-
tion networks [29], we contend that geopolitical factors are also important. Moorthy and
Brathwaite [30] and Jackson and Atkinson [31] show that power competition in the form of
rivalry has a significant positive effect on refugee flows. These networks of animosity may treat
refugee flows as valuable foreign policy tools [2, 32]. On the other hand, networks of amity
may show refugees where they will find welcoming hosts.
Data and methods
Quantitative network analysis can help disentangle which of these networks are important
drivers of refugee flows. We estimate one Multiple Regression- Quadratic Assignment Proce-
dure (MR-QAP) model for each year from 1991–2016 in order to identify the variables that
are significantly related with refugee flows. MR-QAP is a dyadic hypothesis testing method
that accounts for dyad-level correlations and interdependencies across dyads, addressing
known weaknesses of traditional regression analysis [14, 33].
MR-QAP models require input variables to be in the form of square matrices. For the
1991–2016 time period, we therefore use 195 x 195 square matrices for each variable. Prior to
2011, these square matrices are smaller. For example, South Sudan became a country in 2011,
so we work with 194 x 194 square matrices for 2010. With these matrices, the first step is to
estimate an ordinary least squares (OLS) regression. Due to concerns about the sensitivity of
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 2 / 13
be interpreted as representing the official policies
either expressed or implied of the Army Research
Office or the US Government. The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript. There was no additional external
funding received for this study.
Competing interests: The authors have declared
that no competing interests exist.
MR-QAP to collinearity, we checked Variance Inflation Factors (VIFs) for all of our models
[34]. In all cases, our models did not show evidence of collinearity.
Then, we use the estimated t-statistic and compare it to a distribution of t-statistics. That
distribution is calculated through 1000 Monte Carlo simulations where rows and columns of
the dependent variable are shuffled, while maintaining row-column combinations. Our qua-
dratic assignment procedure uses the double semi-partialing permutation method, which is
the most robust permutation approach [34]. This process isolates dyad-level correlations. It is
then possible to estimate a p-value comparing the observed t-statistic with the simulated distri-
bution of t-statistics [35, 36]. We use the netlm function from the statnet package in R for this
analysis [37]. Since the software computes the MR-QAP-adjusted p-value but not an adjusted
t-statistic, we use the adjusted p-values and degrees of freedom for each year to calculate our
own MR-QAP-adjusted t-statistics.
We save the MR-QAP-adjusted t-statistics and OLS t-statistics and then analyze the annual
variation in those t-statistics. For the MR-QAP-adjusted t-statistics, we use k-means clustering
analysis to classify our variables into 3 groups and our years into 3 groups. The clustering by
variables compares the annual variation in MR-QAP-adjusted t-statistics across years. By con-
trast, the clustering by years compares the annual variation in MR-QAP-adjusted t-statistics
across variables. We estimate our clusters from distance matrices based on Pearson’s correla-
tion coefficients. Then, we use multi-dimensional scaling (MDS) analysis to visualize these
classifications.
Table 1 displays the variables and data sources used for the analysis. Replication materials
will be made available online. Our dependent variable, Refugee Flows, comes from refugee data
from the United Nations High Commissioner for Refugees (UNHCR). This resource includes
Table 1. Variable descriptions.
Variable Description Source
Dependent Variable
Refugee Flows First difference in dyadic refugee stocks (Flows
from country i to country j)
UNHCR
IndependentVariables
Security
Gradient
Difference in political terror between destination
and origin country (Gradient)
Political Terror Scale
Wage Gradient Destination country GDP per capita minus origin
country GDP per capita (Gradient)
World Development Indicators
Immigrant
Population
Total migrant stock in 1990, 2000, or 2010 (Flows
from country i to country j)
World Bank Bilateral Migration Matrix
Prior refugee
flows
One year lag of refugee flows (Flows from country i
to country j)
UNHCR
International
trade
Magnitude of international trade (Flows from
country i to country j)
Correlates of War
Regime Type
Gradient
Difference between destination and origin country
regime type (Gradient)
V-DEM
Alliance Existence of an alliance of mutual defense
(Undirected dyad)
ATOP version 4.01
Arms Flows Quantity of arms flows in SIPRI units (Flows from
country j to country i)
SIPRI Arms Transfers Database
Strategic Rivalry Existence of strategic rivalry (Undirected dyad) Thompson & Dreyer (2011); updated
by authors through 2016
Contiguity Dichotomous indicator of land or water contiguity
(Undirected dyad)
Correlates of War
https://doi.org/10.1371/journal.pone.0245712.t001
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 3 / 13
dyadic refugee stocks from 1975–2016, excluding Palestinian refugees since they fall under the
mandate of the United Nations Relief and Works Agency (UNRWA). Before 1990, a large por-
tion of UNHCR’s refugee data is not actually dyadic, so we would advise researchers to only
use UNHCR’s monadic data if they wish to conduct analysis on refugee population data pre-
1990. We calculate the first difference of refugee stocks to obtain net refugee flows. We then
replaced missing values with zero. We also followed existing standard practice and replaced
negative net refugee flow values with zero. Our measure of refugee flows should be considered
as an undercount of flows, since it does not include short-term movements or return move-
ments. This is, however, the best available measurement of refugee flows.
Next, we add our independent variables. While we only discuss the results for full models
with all independent variables, bivariate regressions yield consistent results. Our variable PriorRefugee Flows is a measurement of net refugee flows in the previous year. Our measurement of
immigration, Immigrant Population, comes from the World Bank’s data on bilateral interna-
tional migration. It measures dyadic immigrant stocks as of 1990, 2000, or 2010. We use the
most recent prior immigrant population measure. We then used Benjamin Graham’s and
Jacob Tucker’s data repository to obtain data on GDP per capita from the World Bank’s
World Development Indicators database (Income Gradient), the Varieties of Democracy
(V-DEM) additive polyarchy measure of regime type (Regime Type Gradient), and the Political
Terror Scale (PTS) index value for each country (Security Gradient) [38]. From these monadic
variables, we created gradient matrices to capture the difference in these values between desti-
nation and origin countries. For Security Gradient, we used values coded from Amnesty Inter-
national reports. When those were missing, we used values coded from State Department
reports. This yielded a variable with zero missing values for 2015. For 2011–2014, we used val-
ues coded from Human Rights Watch reports when there were missings for Amnesty Interna-
tional and the State Department. For remaining missing values, we imputed a value of zero.
For Regime Type Gradient, we replaced missing dyad gradient values with zero. For IncomeGradient, we also replaced missing dyad gradient values with zero.
Our Security Gradient and Income Gradient variables only go through 2015, so we used
those values for our 2016 models as well. Our measurement of dyadic arms flows, Arms Flows,comes from annual dyadic data from the Stockholm International Peace Research Institute.
We replaced missing values of Arms Flows with zero. Our measure of international trade,
Trade, comes from the Correlates of War project [39]. This variable only goes through 2014,
so we used the 2014 values for models of refugee flows in 2015 and 2016.
Our measure of contiguity (Contiguity) also comes from the Correlates of War project [40].
Contiguity is a critical geographic variable because it captures the increased familiarity that
people have with neighboring countries, higher likelihood of shared languages and customs,
and easier and cheaper transportation options [29, 41, 42].
Our measure of alliances of mutual defense (Alliance of Defense) comes from the Alliance
Treaty Obligations and Provisions Project (ATOP) [43]. Our rivalry measure, Strategic Rivalry,
comes from the rivalry dataset created by William R. Thompson [16, 44]. The strategic rivalry
dataset codes rivalries through 2010, so we updated the coding of strategic rivalries through
2016. Details about the updated coding of strategic rivalries will be posted on the lead author’s
personal website.
Results
Fig 1 describes post-Cold War refugee flows. At the end of the Cold War, the breakup of the
Soviet Union and Yugoslavia, collapse of the Somali state, and other conflicts drove the world’s
refugee population over 15 million. As the 1990s proceeded though, conflicts progressively
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 4 / 13
ended. The loss of sponsorship for opposing sides from the United States and Soviet Union
removed fuel that had been allowing many wars to persist [45]. This period of conflict termina-
tion coincided with a decline in the world’s refugee population. Fig 1 shows that the global
refugee population progressively fell from 1990 to 2005. In addition to the global refugee popu-
lation falling as conflicts terminate, the refugee population also increased when new conflicts
Fig 1. World refugee population, excluding Palestinian refugees (top) & refugee flows, correlated across years
(bottom).
https://doi.org/10.1371/journal.pone.0245712.g001
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 5 / 13
began. From 2005–2007, the world’s refugee population increased, before remaining relatively
constant until 2012. During the 2012–2016 time period, the world refugee population under
UNHCR’s mandate increased substantially, from roughly 10.5 million to 17 million refugees.
This explosion of refugee flows corresponds with the violent aftermath of the Arab Spring,
which included the fall of governments in Tunisia, Egypt, Yemen, and Libya, as well as civil
wars in Syria, Libya, and Yemen [27]. These developments destabilized other countries not
directly involved in the Arab Spring as well. Libya’s collapse flooded the African Sahel with
fighters and weapons [46]. Syria’s civil war contributed to civil war in Iraq [47]. These chal-
lenges compounded existing issues with the founding of South Sudan in 2011, Somalia’s 2011
drought that triggered more refugee flows than Somalia had produced since the early 1990s,
and the straining of host communities in Iran and Pakistan to host millions of Afghan refu-
gees. Political instability did not really start affecting refugee flows until 2012 because its main
effect initially was to fuel internal displacement in 2011. Beginning in 2012, however, the new
phase of conflict and migration activity increased refugee flows.
The increased activity in the global refugee flow network from 2012–2016 created substan-
tial international concern. In addition to a rapidly growing worldwide refugee population, the
world also experienced refugee flows through different sets of directed dyads. In Fig 1, we illus-
trate this with a correlation network. Our correlation network defines weighted ties between
each year’s set of directed dyads based on the Pearson’s correlation coefficient. For clarity in
our visualization, we exclude ties between years that have a correlation coefficient less than 0.2.
As Fig 1 shows, all correlation coefficients were greater than 0.2. Refugee flows from 2012–
2016 were highly correlated at the directed dyad level (with 2014–2015 having the highest cor-
relation coefficient of 0.86), whereas all other years except for 2007 and 2008 formed a large
and indistinct jumble. These trends in the aggregate and at the directed dyad level suggests
that something new was happening. The 2012–2016 time period was a new phase in global ref-
ugee flow patterns.
Fig 2 displays MDS plots of the k-means clustering analysis that build on this descriptive
insight. Cluster analysis of variables shows that they fit together into three categories: 1) Strate-gic Rivalry, Security Gradient, and Trade; 2) Alliance, Arms Flows, and Contiguity; and 3) PriorRefugee Flows, Immigrant Population, Wage Gradient, and Democracy Gradient. These group-
ings align with the three categories that we discussed in the Introduction: migration networks,
networks of animosity, and networks of amity. Here, there is some support for the view that
Wage Gradient and Democracy Gradient matter for the creation of migration networks, and
then Prior Refugee Flows and Immigrant Population matter for the maintenance of migration
networks [48–50]. Then, Arms Flows and Alliance fit well within the category of networks of
amity. The inclusion of Contiguity in this category suggests that contiguity increases refugee
flows via a “familiarity breeds friendship” effect, rather than by contributing to the creation or
maintenance of migration networks through a “familiarity breeds contempt” effect.
The networks of animosity category functions differently from the networks of amity cate-
gory, supporting the view that friendship and hostility are alternative pathways in international
relations [51]. Fig 3 shows that rivalry becomes more important in driving refugee flows when
trade becomes less important and relative safety becomes more important. Readers should
remember when interpreting Fig 4 that due to how Security Gradient is coded, negative t-sta-
tistics actually mean that there are higher refugee flows when the destination country is safer
than the origin country.
Meanwhile, there are some years where refugee flows actually tend to move towards coun-
tries that are less safe than the origin country (e.g.- 1993, 2003, 2011). Trade generally has a
positive and significant relationship with refugee flows, but its clustering into the “networks of
animosity” category is unexpected. For most of the 1991–2016 time period, Trade has a fairly
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 6 / 13
Fig 2. MDS variables (top) & years (bottom).
https://doi.org/10.1371/journal.pone.0245712.g002
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 7 / 13
constant relationship with refugee flows at around +2. Then, from 2013–2016 it fluctuates in
the opposite direction from Strategic Rivalry.
Fig 3 also displays the t-statistics for migration networks. It is clear that Prior Refugee Flowshas t-statistics that are substantially larger than Immigrant Population, Wage Gradient, and
Democracy Gradient. Immigrant Population also tends to have significant and positive rela-
tionships with Refugee Flows. Wage Gradient is generally not significant. Democracy Gradienthas a positive and significant relationship with refugee flows in some years, but it also does
not generally appear to be important. Moreover, the inclusion of Democracy Gradient in the
“migration networks” category suggests that this variable matters for its role in facilitating the
creation of international migration networks. Like the “networks of animosity” category, the
“migration networks” category has an important shift during the 2013–2016 time period.
Here, the t-statistics for Prior Refugee Flows and Immigrant Population fluctuate in clear oppo-
site directions from 2013–2016.
The third set of t-statistics in Fig 3 is for networks of amity. Alliance and Arms Flows are
usually not statistically significant, but there are several years where Alliance has a positive and
significant relationship with Refugee Flows. Contiguity, meanwhile, is positive and significant
for every year except 2015. Again, 2013–2016 exhibits some different dynamics. Here, Allianceand Contiguity suddenly fluctuate in opposite directions, with Contiguity’s t-statistic generally
shrinking as Alliance’s t-statistic grows. This suggests that there may be a saturation effect in
contiguous states that led to onward migration, which was influenced by geopolitical
considerations.
Fig 3. MR-QAP-adjusted t-statistics networks of animosity (top left); networks of amity (top right); migration networks (bottom);
t-statistic reference lines at +/- 2.
https://doi.org/10.1371/journal.pone.0245712.g003
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 8 / 13
There are important temporal dynamics for Contiguity and Strategic Rivalry that merit
additional analysis. For this purpose, we refer to Fig 4. All plots include a fitted line in order
to clarify the temporal trend. For both variables, we include one plot with the time series of
MR-QAP-adjusted t-statistics and one plot with the time series of OLS t-statistics. From these
plots, it appears that Strategic Rivalry has a growing effect over time in the OLS results, whereas
its effect is relatively constant over time in the MR-QAP results. Contiguity does not have nota-
ble differences between the t-statistics from MR-QAP and OLS. This distinction suggests that
Strategic Rivalry may influence the structure of the global refugee flow network, not just the
magnitude of individual refugee flows. Our results do not allow us to specify the specific net-
work structures in the global refugee flow network that are influenced by strategic rivalry, but
the observation that network structures may be influenced by dyadic relationships is important
in and of itself. Meanwhile, Contiguity has a declining effect on Refugee Flows over time. In
other words, since the end of the Cold War, refugees have been moving further away from
their origin countries.
We also examine our observation that the drivers of refugee flows appear to shift during the
2013–2016 time period. In Fig 1, we observed that the directed dyads of the global refugee flow
network are strongly correlated from 2012–2016. This is also a period where the global refugee
population rose substantially. Therefore, the world was experiencing a surge in refugee flows
and that surge was occurring through a new set of directed dyads. Fig 2 shows that when we
cluster t-statistics over years, that surge was also being driven by a new set of factors. When t-
statistics for the 1991–2016 models are clustered into three groups over years, there is a lot of
Fig 4. Rivalry & contiguity with fitted lines (MR-QAP results top-left and bottom-left, OLS results top-right and bottom-right).
https://doi.org/10.1371/journal.pone.0245712.g004
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 9 / 13
noise. One important distinction is that the years 2010–2012 cluster into different groups than
the years 2013–2016. Our year clusters do not just reflect when refugee populations were ris-
ing, falling, or flat, since there is not a coherent pattern for how years cluster before 2013 in Fig
2. These observations all indicate that something important changed starting in 2012. We can-
not specify one trigger with certainty, but as we discussed earlier, the Arab Spring and related
instability in 2011 is a plausible culprit. If the Arab Spring were actually unrelated to this shift
in refugee flow patterns, our finding remains that something changed during the early 2010s.
Discussion
In this paper, we have demonstrated the value of explaining refugee flows as a multi-layer net-
work. An analysis that incorporates multiple overlapping networks allows us to account for
many kinds of network dependencies. Our analysis describes the evolution of the global refu-
gee flow network since the end of the Cold War. Considering correlations between the sets of
existing refugee flow directed dyads in each year, we observe a noisy process from year to year.
The 2012–2016 time period, however, stands out as a period with highly inter-correlated refu-
gee flows across years. We applied MR-QAP and learned that refugee flows vary based on
variation in factors related to international migration networks, networks of animosity, and
networks of amity.
There are several possibilities for future research. It would be valuable to use these macro-
level findings to guide further research on refugee flow ego networks (e.g., focus on a single
country’s flows to unpack spatial and temporal dynamics) and on emerging drivers of refugee
flows like climate change. Like the climate-conflict research program [52], climate-induced
migration research argues that the role of climate and environmental factors is likely to increase
substantially in the coming decades. In addition, we find evidence for a role of inter-state amity,
as well as animosity, in facilitating refugee flows. New research should follow-up on our find-
ings and consider the role of alliances and rivalries. Since rivalry may influence network struc-
tures of the global refugee network, not just dyadic refugee flows, future research could also
explore which network structures in the refugee flow network change due to rivalry. Finally, the
shift in refugee flow patterns after 2011 suggests that additional research on the Arab Spring
and other key events from the early 2010s could contribute to broad understandings of changes
in the international system, beyond changes in the Middle East alone.
Supporting information
S1 File.
(ZIP)
Acknowledgments
The authors acknowledge support from the Democratic Statecraft Lab at the University of Vir-
ginia and from research assistants at the University of Florida (Elise Geissler and Isaac “Rico”
Mirti). We also acknowledge feedback from Rachata Muneepeerakul, Rafael Muñoz-Carpena,
Upmanu Lall, David Griffith, and Michael J. Puma.
Author Contributions
Conceptualization: Justin Schon, Jeffrey C. Johnson.
Data curation: Justin Schon.
Formal analysis: Justin Schon.
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 10 / 13
Funding acquisition: Jeffrey C. Johnson.
Investigation: Justin Schon.
Methodology: Justin Schon.
Project administration: Jeffrey C. Johnson.
Resources: Justin Schon.
Software: Justin Schon.
Supervision: Jeffrey C. Johnson.
Validation: Justin Schon.
Visualization: Justin Schon.
Writing – original draft: Justin Schon.
Writing – review & editing: Justin Schon.
References
1. Salehyan I, Gleditsch KS. Refugees and the Spread of Civil War. International Organization.
2006:335–66.
2. Betts A, Loescher G. Refugees in international relations: Oxford University Press; 2011.
3. Beine M, Docquier F, Rapoport H. Brain drain and economic growth: theory and evidence. Journal of
development economics. 2001; 64(1):275–89.
4. Choi S-W, Salehyan I. No good deed goes unpunished: refugees, humanitarian aid, and terrorism.
Conflict Management and Peace Science. 2013; 30(1):53–75.
5. Ruegger S. Refugees, ethnic power relations, and civil conflict in the country of asylum. Journal of
Peace Research. 2019; 56(1):42–57.
6. Triandafyllidou A, Maroukis T. Migrant Smuggling: Irregular Migration from Asia and Africa to Europe.
New York: Palgrave Macmillan; 2012.
7. Longo M. The politics of borders: Sovereignty, security, and the citizen after 9/11: Cambridge University
Press; 2017.
8. Czaika M, Haas Hd. The Globalization of Migration: Has the World Become More Migratory? Interna-
tional Migration Review. 2014; 48(2):283–323.
9. Windzio M. The network of global migration 1990–2013: Using ERGMs to test theories of migration
between countries. Social Networks. 2018; 53:20–9.
10. Maoz Z. Networks of Nations: The Evolution, Structure, and Impact of International Networks, 1816–
2001. Cambridge: Cambridge University Press; 2010.
11. Zhukov YM, Stewart BM. Choosing Your Neighbors: Networks of Diffusion in International Relations.
International Studies Quarterly. 2013; 57(2):271–87.
12. Cranmer SJ, Menninga EJ, Mucha PJ. Kantian fractionalization predicts the conflict propensity of the
international system. Proceedings of the National Academy of Sciences. 2015; 112(38):11812–6.
https://doi.org/10.1073/pnas.1509423112 PMID: 26338977
13. Paniagua J, Peiro-Palomino J, Picazo-Tadeo AJ. Asylum migration in OECD countries: In search of lost
well-being. Department of Applied Economics II, Universidad de Valencia; 2020.
14. Oatley T, Winecoff WK, Danzman SB, Pennock A. The Political Economy of Global Finance: A Network
Model. Perspectives on Politics. 2013; 11(1):133–53.
15. Boulding K. Conflict and Defense: A General Theory. New York: Harper & Row; 1962.
16. Thompson W. Identifying Rivals and Rivalries in World Politics. International Studies Quarterly. 2001;
45:557–86.
17. Fagiolo G, Mastrorillo M. Does human migration affect international trade? A complex-network per-
spective. PloS one. 2014; 9(5):e97331. https://doi.org/10.1371/journal.pone.0097331 PMID:
24828376
18. Schon J. Motivation and opportunity for conflict-induced migration: An analysis of Syrian migration tim-
ing. Journal of Peace Research. 2019; 56(1):12–27.
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 11 / 13
19. Leblang D. Harnessing the diaspora: dual citizenship, migrant return remittances. Comparative Political
Studies. 2017; 50(1):75–101.
20. Orchard P. A Right to Flee: Refugees, States, and the Construction of International Society. Cam-
bridge, UK: Cambridge University Press; 2014.
21. Landau L. The Humanitarian Hangover: Displacement, Aid, and Transformation in Western Tanzania.
Johannesburg: Witwatersrand University Press; 2008.
22. Whitaker B. Changing Priorities in Refugee Protection: The Rwandan Repatriation from Tanzania.
Problems of Protection: The UNHCR, Refugees, and Human Rights. New York: Routledge; 2003.
p. 141–54.
23. Verdirame G. Human Rights and Refugees: The Case of Kenya. Journal of Refugee Studies. 1999; 12
(1):54–77.
24. Carter DB, Poast P. Why do states build walls? Political economy, security, and border stability. Journal
of Conflict Resolution. 2017; 61(2):239–70.
25. Hassner RE, Wittenberg J. Barriers to Entry: Who Builds Fortified Boundaries and Why? International
Security. 2015; 40(1):157–90.
26. Avdan N, Gelpi CF. Do Good Fences Make Good Neighbors? Border Barriers and the Transnational
Flow of Terrorist Violence. International Studies Quarterly. 2016; 61(1):14–27.
27. Lynch M. The Arab uprisings explained: New contentious politics in the Middle East: Columbia Univer-
sity Press; 2014.
28. Lacher W. The Malian crisis and the challenge of regional cooperation. Stability: International Journal of
Security and Development. 2013; 2(2).
29. Pedersen PJ, Pytlikova M, Smith N. Selection and network effects—Migration flows into OECD coun-
tries 1990–2000. European Economic Review. 2008; 52(7):1160–86.
30. Moorthy S, Brathwaite R. Refugees and rivals: The international dynamics of refugee flows. Conflict
Management and Peace Science. 2016:0738894216657047.
31. Jackson JL, Atkinson DB. The Refugee of My Enemy Is My Friend: Rivalry Type and Refugee Admis-
sion. Political Research Quarterly. 2019; 72(1):63–74.
32. Greenhill K. Weapons of Mass Migration: Forced Displacement, Coercion, and Foreign Policy. Ithaca,
New York: Cornell University Press; 2010.
33. Cranmer SJ, Leifeld P, McClurg SD, Rolfe M. Navigating the range of statistical tools for inferential net-
work analysis. American Journal of Political Science. 2017; 61(1):237–51.
34. Dekker D, Krackhardt D, Snijders TA. Sensitivity of MRQAP tests to collinearity and autocorrelation condi-
tions. Psychometrika. 2007; 72(4):563–81. https://doi.org/10.1007/s11336-007-9016-1 PMID: 20084106
35. Krackardt D. QAP partialling as a test of spuriousness. Social networks. 1987; 9(2):171–86.
36. Krackhardt D. Predicting with networks: Nonparametric multiple regression analysis of dyadic data.
Social networks. 1988; 10(4):359–81.
37. Handcock MS, Hunter DR, Butts CT, Goodreau SM, Morris M. statnet: Software tools for the Statistical
Modeling of Network Data 2003 [http://statnet.org.
38. Graham BA, Tucker JR. The international political economy data resource. The Review of International
Organizations. 2019; 14(1):149–61.
39. Barbieri K, Keshk OM, Pollins BM. Trading data: Evaluating our assumptions and coding rules. Conflict
Management and Peace Science. 2009; 26(5):471–91.
40. Douglas S, Tir J, Philip S, Charles G. The Correlates of War Project Direct Contiguity Data. Conflict
Management and Peace Science. 2002; 19(2):59–68.
41. Huang RR. Distance and trade: Disentangling unfamiliarity effects and transport cost effects. European
Economic Review. 2007; 51(1):161–81.
42. Okour Y. Understanding the transportation experiences of refugees: A case study of Iraqi and Syrian
refugees in Erie County, NY: State University of New York at Buffalo; 2019.
43. Leeds B, Ritter J, Mitchell S, Long A. Alliance treaty obligations and provisions, 1815–1944. Interna-
tional Interactions. 2002; 28(3):237–60.
44. Thompson W, Dreyer D. Handbook of International Rivalries: 1494–2010. Washington, DC: CQ Press;
2011.
45. Kalyvas S, Balcells L. International System and Technologies of Rebellion: How the End of the Cold
War Shaped Internal Conflict. American Political Science Review. 2010; 104(3):415–29.
46. Conflict Armament Research. Investigating Cross-border Weapon Transfers in the Sahel. London;
2016.
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 12 / 13
47. Lister CR. The Syrian Jihad: Al-Qaeda, the Islamic State and the Evolution of an Insurgency: Oxford
University Press; 2016.
48. Massey DS, Arango J, Hugo G, Pellegrino A, Taylor JE. Theories of International Migration: A Review
and Appraisal. Population and Development Review. 1993; 19:431–66.
49. Stark O, Bloom DE. The new economics of labor migration. The American Economic review. 1985; 75
(2):173–8.
50. Menjıvar C. Fragmented ties: Salvadoran immigrant networks in America: Univ of California Press;
2000.
51. Morrow JD. Arms versus Allies: Trade-offs in the Search for Security. International Organization.
1993:207–33.
52. Mach KJ, Kraan CM, Adger WN, Buhaug H, Burke M, Fearon JD, et al. Climate as a risk factor for
armed conflict. Nature. 2019: 1. https://doi.org/10.1038/s41586-019-1300-6 PMID: 31189956
PLOS ONE How inter-state amity and animosity complement migration networks to drive refugee flows
PLOS ONE | https://doi.org/10.1371/journal.pone.0245712 January 27, 2021 13 / 13
Recommended